Reinforcement Learning x World Model: Brain Surgery

We use video LLMs as an agent to roam through World Models and receive real-time reward signals that guides LLM's next action to reach our goal.